citygoods: tools for modeling urban freight distribution

CityGoods: Tools for modeling urban freight

distribution1st Scientific and Technical workshop

Bologna, Italy5/11/2013

Presented by:Daniele Vigo – DEI University of

Bologna

Talk outline

• Backround and Motivation: – CityPorts and Merope Interreg III/B projects– Emilia-Romagna’s towns logistic surveys

• A unified modeling framework for Urban Logistics (CityGoods)– Step 1: a demand generation model– Step 2-3: distribution and assignment models– Step 4: GityGoods modeling suite

• Conclusions

2CityGoods: Daniele VigoBologna November 5, 2013

The CityPorts Project

• CityPorts (EU INTERREG III/B 2003-05), coordinated by REE, proposed a general methodological framework for the design ad evaluation of City Logistics Actions (support initiatives):– Infrastructures– Policies and regulations …

• analysis of the different supply chains and their impact on the different zones of the urban area

• Another twin project, Merope, involved other towns of the region with similar objectives


The Zones-Supply Chains Grid

• Logistic characterization of towns is based on the construction of the Zones-Supply Chains (ZS) Grid as a “reading guide” of the town in Logistic terms

4

measure of criticity:

• qualitative (█, █, █)

• quantitative (n. of opns)Supply Chain

Zon

es

CityGoods: Daniele VigoBologna November 5, 2013

Logistic Actions and ZS Grid

• A Logistic Action (Infrastructure, Policy) may be mapped into the ZS Grid to evaluate it effectiveness

5

Supply Chain

Zon

es

Non SC-specialized DC for a specific Zone

SC-specialized DC for several Zones

Expected benefit of the Action on the ZS element


Logistic Actions Evaluation

• Action mapping into a ZS Grid gives the basis for Action effect evaluation (and possible re-design)

• Need of a quantitative ZS Grid• Need of a Supply-Chain-based Demand

Generation Model – to define the ZS Grid – to be used within classical transp. models


RER City logistic surveys• In the years 2003-05 Emilia-Romagna performed an extensive

survey of City Logistics phenomenon for all ER towns– CityPorts, Merope, Regional programs …

• Huge and fine-grained data source– Quite homogeneous (CityPorts survey model)– 3 main surveys:

• Demand Generation (interviews at shops…), • Demand Attraction (interviews at logistic operators)• Flows/operations (interviews to vehicles)

– 11 towns (all with > 50K inhabitants), thousands of interviews• A unique modeling opportunity !


CityGoods Modeling framework

• Unified modeling approach:– Description of City Logistics phenomena

• For a specific town and for a regional territory– Definition of qualitative and quantitative indicators of City Logistics in

ER towns– That may be used for evaluation and planning purposes

• CityPorts methodology and Classical transportation analysis

• Classical modeling framework (Generation, Distribution, Assignment)

• Demand Generation exploits the hierarchical structure of activity classification systems

• Distribution takes into account that vehicles can perform more than one stop in a tour


CityGoods history

• The prototypes of the models were developed in 2004 for RER

• In 2005-06 models are validated on real-world data from RER surveys

• In 2006-07 the models were integrated into a unified package (CityGoods) used by RER

• CityGoods was developed by Guido Gentile and Daniele Vigo and is distributed by SISTEMA


Demand Generation Models• General survey: Russo & Comi (2004)

– gravitational, 4 Phases (Hutchinson 1974, Odgen 1992, List & Turnquist 1994, Taylor 1997, He & Crainic 1998, Gorys & Hausmanis 1999 …): more suited to a urban scale

– input-output (Harris & Liu, 1998)– spatial price equilibrium (Oppenheim 1994, Nagurney, 2002)– CityGoods: (Gentile & Vigo, European Transport, 2013)

• Some problems:– Generation:

• intrinsic approximation introduced by aggregating many economic activities into few categories

• a given economic activity generates movements belonging to different Supply Chains

– Distribution: a vehicle performs many deliveries/pickups in a tour


Other experiences

• FRETURB (L.E.T., Lyon, Fr): general model for the evaluation of the impact of Logistic Actions– Based on 3 detailed surveys on French towns

(Marseilles, Bordeaux and Dijon)– Regression-based model – Software tool distributed by French Ministry of

Transport to all French Municipality• VISEVA (Friedrich et al 2003), Good Trips

(Boerkamps, 1999)


Demand Generation Model

• Objective: estimate the yearly number of operations generated by each SC - Zone

• Starting Points:– ER surveys on Demand Generators:

• Small samples wrt Universe (e.g. Bologna: 250-500 out of 35131)

• Rich of logistic information (operations generated per SC, time distribution, type of vehicles …)

– Universe• Municipality, CCIAA Data … ASIA ER Database• No Supply-chain related classification (only ATECO/NACE

economic classification, NAICS in USA …)


Demand Generation Model (2)

• Overall approach:– No “a-priori” aggregation of activities into categories (SC)– Characterize demand generation directly using the

ATECO/NACE classification of the generators (operations per NACE code)

– Hundreds of codes and small samples! Exploit the hierarchic structure of the classification within the model

– Use survey data to calibrate the model and define the specific SC generation models

• Result:– Very fine-grained information wrt to classical index-by-

category approaches


ATECO-NACE tree

• 5 Digits code with hierarchic structure


ATECO-NACE tree (2)

• Mapping of the Universe into the NACE tree gives immediate indicators of town structure (overall and spatial=per Zone)

15

Bologna Universe


NACE-Based Model

• Main assumption– The n. of operations generated by a specific NACE

code (e.g. 502 Vehicle Maintenance) should take into account:

• Those generated by the “descendant” codes (5020, 5021, … 50201,…,50205)

• Those generated by “parent” classes (50, 5)– Measured by two contributes:

• the relative weight (n. of elements in the Universe) of the subtree rooted at the code

• relative weight of the path to the tree root


Model Formulation

• NACE tree structure:– N set of NACE codes– f(i) father of code iN– FS(j) = {iN: f(i) = j} jN– r root of the tree

• Mis number of yearly operations of the supply chain sS generated by code iN (model output)

• mis number of yearly operations associated to the link entering iN (parameters to be determined by calibration)

17

j=f(i)

i

FS(j)


Model Formulation (2)

• Mis = Wis + His iN

• Wis contrib. to Mi of the subtree rooted in i

• His contrib. to Mi of the path from i to r (for leaves Mis = His)

• His = mis + Hf(i)s iN (computed in topological order)

• Wis = jFS(i) j (Wjs + mjs) iN (computed in reverse topological order)

i probability that the child of f(i) is iN (computed statistically from the Universe)


NACE-Based Model (2)

• The overall model defines the total number of operations of a SC per year generated as a function of:– the NACE code – the n. of employees in each local unit

• Survey data are used to– calibrate the model – obtain the distribution of the different attributes

(parking type, time of service…) for each SC


NACE-Based Model (3)

• A spatial model is derived by using – user-defined Zones (Cityports Macro-Zones,

Transportation studies zones …)– distribution of the Universe in the Zones

• through geocoding by using a commercial street network (Navteq) available for all towns

• Municipality-owned GIS …


CityGoods Overview

• CityGoods is a GIS application based on an ACCESS-like DBMS …


CityGoods Overview

• … with specific tools to model CityLogisitcs


CityGoods Overview• It imports the street network (e.g. NavTeq shape file)


CityGoods Overview• It imports the City Zones (shape file)


CityGoods Overview

• Geocodes the survey data …


CityGoods Overview• Geocodes the Universe data …


CityGoods Overview• Imports the surveys to obtain the movements per SC and

other logistics attributes


CityGoods Overview• Applies the generation model to obtain the number of

movements per NACE Code and per year …


CityGoods Overview• … and produces the ZS matrix …


CityGoods Overview• Distribution model uses Logistic Portals (Geocoded)…


CityGoods Overview• Computes all the distances btw inner points and Portals


CityGoods Overview

• Distribution: Gravitational model modified to account for multiple deliveries within each Zone and applied for each supply chain


CityGoods Overviewz

• … produces the OD matrices per supply chain that are assigned to the network (possibly jointly with people movements if available)


CityGoods Overview

• … obtaining the freight vehicles flows


CityGoods Overview• … data may be easily exported to be used by other

applications (e.g. VISUM)


Typical Use-Cases

• Construction of the specific model for a given town using “its” surveys

• Construction of the “regional” model using all the surveys (often covering different SCs)

• Application of the town or regional model to other towns (without additional surveys)


Data requirements

• Universe:– Address, ATECO/NACE code, (employees)

• Street Network:– shape file with street network (tested with

Navteq but works also with others)• Zones

– shape file with geocoded zone boundaries• Supply chains

– Supply Chain list, (other distribution parameters)


38

• Sample:– Address, ATECO/NACE code, (employees),

movements per SC, (other logistic infos)• ATECO/NACE tree• Logistic Portals

– shape file with geocoded portals


39

Conclusions …

• Very effective modeling tool– “Soft” data requirements:

• Simple Universe data, NACE tree, commercial street graphs

– Fine granularity of results and excellent quality of real-world testing

– Easy portability of the model to different towns with/without specific additional surveys


Thanks you for your attention

•Daniele VigoDEI University of [email protected]

•Guido GentileDICEA University of Rome “La Sapienza”[email protected]


mailto:[email protected]

mailto:[email protected]

citygoods: tools for modeling urban freight distribution

Documents

daniele vigobologna

daniele vigo bologna

logistic terms

zonessupply chains zs

demand generation interviews

rer city logistic surveysin

modeling urban freight

demand generation modelstep